Predictors of Rapid Brain Imaging in Acute Stroke
Analysis of the Get With The Guidelines–Stroke Program
Background and Purpose—National guidelines recommend patients with acute stroke undergo brain imaging within 25 minutes of emergency department arrival. Delayed brain imaging may reduce the effectiveness of thrombolysis or render patients ineligible.
Methods—Data from the Get With The Guidelines–Stroke program from 2003 to 2009 were analyzed to determine overall imaging rates, temporal trends, and predictive variables associated with door-to-imaging times in patients who presented to an emergency department within 2 hours of stroke symptom onset and did not have clear reasons for withholding thrombolysis. Multivariable logistic regression adjusting for within-hospital clustering was performed to identify independent predictors of brain imaging within 25 minutes of emergency department arrival.
Results—Brain imaging was performed within 25 minutes in 41.7% of patients. Rates of imaging within 25 minutes increased from 2003 to 2009 (33.3%–44.5%). In the multivariable model, the following variables were associated with less likelihood of imaging being completed within 25 minutes: age >70 years; female gender; nonwhite race; history of diabetes, peripheral vascular disease, or prosthetic heart valve; transportation other than ambulance; arrival >60 minutes after symptom onset; and hospital location in the Northeast. Patients with National Institutes of Health Stroke Scale scores of 16 to 25 (compared with other strata) were most likely to have imaging completed within 25 minutes.
Conclusions—Most patients with acute stroke symptoms do not have brain imaging performed within the recommended 25 minutes. Future quality improvement initiatives should focus on reducing door-to-imaging times with a specific emphasis on the predictive variables identified in this analysis.
Treatment with intravenous thrombolysis (tPA) has been shown to improve outcomes after stroke, both in clinical trials1 and clinical practice.2,3 Earlier treatment with tPA is associated with higher rates of good clinical outcomes.4,5 The desire for earlier treatment has led to the development of time-based guidelines for the emergency evaluation of patients with acute stroke, including a recommendation that brain imaging be completed within 25 minutes of emergency department arrival.6,7
Previous studies have investigated potential reasons for delays in the evaluation and treatment of patients with acute stroke. Transport by emergency medical services (EMS),8–10 advanced hospital notification by EMS,11 shorter symptom onset-to-presentation times,12,13 and higher stroke severity (as measured by the National Institutes of Health Stroke Scale [NIHSS] score)12 have been associated with more rapid emergency evaluation, including imaging.
The purpose of this study was to examine adherence to the door-to-imaging benchmark of 25 minutes for patients with suspected stroke. We sought to identify trends over time as well as perform a comprehensive analysis of patient- and hospital-level characteristics that are associated with rapid brain imaging.
Characteristics of Get With The Guidelines–Stroke (GWTG) have been previously described.14,15 Participating hospitals use an Internet-based platform to enter data, receive decision support, and obtain feedback through on-demand reports of performance on quality measures on consecutive stroke and transient ischemic attack admissions. Trained hospital personnel abstract data using standardized definitions and coding instructions. Data quality is monitored for completeness/accuracy. Hospital characteristics were based on American Hospital Association data.16 Medical history was defined based on pre-existing conditions. Participating hospitals receive approval to enroll cases without individual patient consent under the common rule or a waiver of authorization and exemption from subsequent review by their Institutional Review Board. Outcome Sciences, Inc serves as the data collection and coordination center for GWTG. The Duke Clinical Research Institute serves as the data analysis center and has Institutional Review Board approval to analyze the data set.
The initial study population (n=639 494) included patients enrolled in GWTG from April 2003 through June 2009 with a primary discharge diagnosis of stroke (all subtypes) whose point of hospital entry was the emergency department (ED). Only patients with a documented symptom onset ≤2 hours before ED arrival were analyzed, because these patients are potentially eligible for tPA within 3 hours according to recommended guidelines of treatment within 60 minutes of ED arrival; this omitted 505 157 patients from analysis. Patients from participating hospitals with <30 patients from 2003 to 2009 were excluded (n=186). The following exclusion criteria were also applied: patients with missing data for any field essential for analysis (eg, ED arrival time; n=6861) and patients receiving tPA at another institution (n=190) or any experimental stroke therapies (n=295). Patients meeting these criteria (n=126 795) were used to report rates of imaging completion within 25 minutes by stroke subtype (ischemic, intracerebral hemorrhage, subarachnoid hemorrhage, transient ischemic attack, stroke not otherwise specified).
We next excluded patients who did not undergo brain imaging (n=1104) and patients who had a tPA contraindication/warning that would be apparent on or shortly after ED arrival, because treating providers might postpone urgent brain imaging if an obvious contraindication existed or attempt to treat reversible contraindications/warnings before imaging (n=59 784). Patients with the following characteristics were therefore excluded: systolic blood pressure >185 mm Hg and/or diastolic blood pressure >110 mm Hg; seizure at symptom onset; surgery/trauma within 14 days; intracranial/spinal surgery, head trauma, or stroke within 3 months; gastrointestinal/internal bleeding within 21 days; any history of intracranial hemorrhage, aneurysm/vascular malformation, or brain tumor; serum glucose <50 mg/dL or >400 mg/dL; increased bleeding risk secondary to comorbid conditions; life expectancy <1 year, severe comorbid illness, or comfort-care measures only on ED arrival; pregnancy; and rapid neurological improvement or stroke symptom severity too mild. This subset (n=65 907) was defined as patients expected to move rapidly to imaging studies and rates of imaging completion within 25 minutes by stroke subtype were again reported.
Patients with ischemic stroke without obvious contraindications to tPA (as listed previously) who would therefore be expected to have rapid brain imaging (n=40 377) were used for univariate and multivariable analyses.
The primary outcome for this analysis was rapid brain imaging, defined as time from ED arrival to completion of imaging (door-to-imaging time [DIT]) of 25 minutes or less. A secondary outcome was DIT considered as a continuous variable.
The association of potential predictor variables with rapid brain imaging was assessed using single-predictor logistic regression models and generalized estimating equations to account for within-hospital clustering of data. Continuous predictor variables were divided at their median values when reporting unadjusted event rates; however, they were entered in continuous form into the multivariable-adjusted logistic regression models. Patient-level variables included demographic data, NIHSS score, medical history, and characteristics of ED arrival. Hospital-level variables included hospital size, volume of patients with stroke, teaching status, and location. All variables had ≤10% missing data except NIHSS (missing in 33.9%), advanced EMS notification (18.3%), and hospital volume of patients with stroke (14.5%). The relationships between these variables and DIT were assessed using Spearman correlation statistics for continuous variables and Wilcoxon rank-sum tests or Kruskal–Wallis tests for categorical variables.
The effect of DIT on other early tPA-related outcomes (frequency of tPA administration, door-to-needle times, and frequency of tPA treatment within 60 minutes) were analyzed using χ2 tests.
A multivariable logistic regression model, using generalized estimating equations to account for within-hospital clustering effects, was used to identify the predictors associated with rapid brain imaging. Predictor variables were considered for model entry if an association between the predictor and rapid imaging was seen in the univariate analysis (probability value <0.10) and data were missing for <10% of patients (to maintain overall sample size). For predictor variables with multiple categories, a probability value of <0.10 for comparison of at least 1 category to the reference group was required for entry. Candidate variables were entered into the model and backward selection used to determine independent predictors using a threshold probability value of <0.05. For variables with multiple levels, each level was compared with the reference group; if a significant difference was not found, the specific level was then combined with the reference group.
In a subgroup analysis among patients with NIHSS score available (n=23 818), the relationship between NIHSS score and rapid brain imaging was assessed using both linear and quadratic terms for NIHSS score in a logistic regression model adjusted for other predictors identified in the full multivariable model.
All statistical analysis was performed using SAS software Version 9.2 (SAS Institute, Cary, NC).
After applying exclusion criteria, 126 795 patients were included in descriptive analyses. Imaging was completed within 25 minutes in 36.8% of all patients with stroke and 41.8% of patients with ischemic stroke (Table 1). After excluding patients with reasons for withholding tPA that might postpone or delay imaging, rates of DIT ≤25 minutes remained low: 41.7% of all patients with stroke and 46.9% of patients with ischemic stroke. Rates of DIT ≤25 minutes in all patients with stroke increased from 2003 to 2009 (33.3% versus 44.5%, P<0.0001 for trend over time; Figure 1). Across hospitals, rates of DIT ≤25 minutes for patients with ischemic stroke ranged from 0% to 100% (median, 42%; interquartile range, 27%–57%).
Table 2 shows the proportion of patients with ischemic stroke undergoing rapid imaging categorized by predictive variables. Older age (≥75 years), female gender, and black race (compared with white) were all associated with lower probability of DIT ≤25 minutes. A known history of smoking was associated with a higher probability of rapid imaging. Known diagnoses of diabetes, hypertension, peripheral vascular disease, or coronary artery disease were associated with lower probability of DIT ≤25 minutes. Patients arriving by means other than EMS were less likely to have DIT ≤25 minutes. Patients whose ED arrival was preceded by prearrival notification and those who arrived to the ED within 60 minutes of symptom onset were more likely to have rapid imaging.
Patients admitted to hospitals in the Northeast (compared with the West) and patients admitted to hospitals with ≤100 stroke discharges/year (compared with hospitals with >300 stroke discharges/year) were less likely to have rapid imaging.
Similar relationships were observed with patient- and hospital-level predictors when DIT was analyzed as a continuous variable (online-only Supplemental Table I; http://stroke.ahajournals.org).
Compared with those with longer imaging times, patients with DIT ≤25 minutes were more likely to be treated with tPA (63.5% versus 38.1%, P<0.0001). Patients with DIT ≤25 minutes also had shorter median door-to-needle times (75 versus 89 minutes, P<0.0001) and were more likely to receive tPA within 60 minutes of arrival (28.5% versus 13.6%, P<0.0001).
The following variables were found to be associated with a lower likelihood of DIT ≤25 minutes in patients with ischemic stroke: age (>70 years); female gender; nonwhite race; history of diabetes, peripheral vascular disease, or prosthetic heart valve; transportation to ED other than by EMS; time between symptom onset and ED arrival ≥60 minutes; and hospital location in the Northeast (Table 3).
The relationship between NIHSS and DIT ≤25 minutes was analyzed in a separate analysis. After adjusting for variables significant in the overall multivariable analysis, NIHSS score was a significant predictor of rapid imaging (P<0.0001). Rates of DIT ≤25 minutes across different strata of NIHSS scores showed an inverted U-shaped response with the highest probability of rapid imaging found in patients with NIHSS scores in the 16 to 20 and 21 to 25 strata (Table 4; Figure 2).
Our results demonstrate that many patients with stroke still experience significant delays between ED arrival and completion of their imaging studies. Specifically, less than half of patients presenting with stroke symptoms have imaging studies completed within the recommended 25-minute timeframe, even among apparently ideal candidates for tPA at the time of ED arrival.
Completion of a brain CT or MRI is a required step before administration of tPA. Because the benefit of tPA decreases as the time from symptom onset increases, delays in brain imaging that result in longer door-to-needle times (as seen in our study) may lead to worse patient outcomes. Target: Stroke17 is an ongoing American Heart Association initiative, which should provide further insights into imaging/treatment barriers. A better understanding of the relationships among DIT, overall door-to-treatment times, and other time-based quality benchmarks is warranted. Other barriers to treatment likely exist, because the median door-to-treatment time was still 75 minutes even when DIT was <25 minutes.
Our findings indicate that specific populations may be at greater risk of delays in brain imaging. Older patients, women, and individuals of nonwhite races were all less likely to receive imaging within 25 minutes. Prior studies have shown differences in stroke-related processes of care for women18,19 and racial minorities.20 Our results suggest that differences in care may occur shortly after ED arrival.
The rate of imaging completion within 25 minutes for diabetic patients is notable. Patients with diabetes are at higher risk of stroke, yet in our multivariable analysis, their adjusted odds of early brain imaging were 15% lower than nondiabetic patients. This difference was found despite the exclusion of patients with critical glucose values on ED arrival, patients whose imaging might be justifiably delayed until these values could be corrected. Similar findings have been observed previously11; 1 possible explanation is that neurological symptoms in diabetic patients with otherwise acceptable glucose values (between 50 and 400 mg/dL) are being attributed to consequences of abnormal glucose metabolism rather than brain ischemia, and acute stroke management pathways are not being activated. Patients with peripheral vascular disease were also less likely to have imaging completed quickly, again possibly related to stroke symptoms being misinterpreted in this population; more focused studies in these patients may be indicated to better understand these results.
Our results corroborate prior studies showing that arrival by ambulance is associated with more rapid assessment, including imaging. In our model, transportation by means other than ambulance was associated with 47% lower adjusted odds of rapid imaging. This illustrates the need for ongoing public education regarding identification of stroke symptoms and immediate EMS activation.21 Advanced hospital notification by EMS was strongly associated with shorter DIT in our univariate results but could not be entered into our multivariable model due to missing data.
We did not find a consistent association between hospital case volume and DIT. One possible explanation may be that high-volume stroke hospitals may also have a higher burden of trauma and other acute disorders that limit resource use.
Compared with other strata, patients with severe strokes (NIHSS 16–25) were most likely to have imaging completed within 25 minutes. Patients with scores >25 may have justifiable delays in imaging for medical reasons such as loss of airway control. NIHSS was missing in one third of patients, reflecting its incomplete use in clinical practice; therefore, results of this subgroup analysis should be interpreted in this context.
There are several limitations to this study. GWTG hospitals are voluntarily participating in a quality improvement program and may not be representative of all US hospitals. Because participating hospitals are interested in improving their acute stroke care, nonparticipating hospitals may have longer DIT than reported here. DIT and other data were abstracted from the medical record and accuracy is dependent on quality of documentation. Other unmeasured confounders may be impacting the timeliness of stroke imaging such as patient preferences/concerns about imaging, which are not captured in our data set.
There is significant room for improvement in the acute evaluation of patients with suspected stroke. Certain patient populations may be at risk for delays in brain imaging, specifically older patients, women, nonwhite patients, and diabetics. Future quality improvement efforts should focus on streamlining the evaluation of patients with acute stroke overall with an emphasis on reducing DIT in these populations. Such efforts have the potential for reductions in door-to-treatment times that may translate into improved patient outcomes.
Sources of Funding
Supported in part by a Get With The Guidelines Young Investigator Award. The GWTG–Stroke program is provided by the American Heart Association and is currently supported in part by a charitable contribution from Ortho-McNeil. Get With The Guidelines–Stroke has been funded in the past through support from Boehringer-Ingelheim, Merck, Bristol-Myers Squibb/Sanofi Pharmaceutical Partnership, and the American Heart Association Pharmaceutical Roundtable.
L.H.S. serves as the unpaid chair of the national steering committee for the GWTG program and as a paid stroke systems consultant for the Massachusetts Department of Public Health.
Natan M. Bornstein, MD, was the Guest Editor for this paper.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.626374/-/DC1.
- Received May 23, 2011.
- Revision received December 31, 2011.
- Accepted January 23, 2012.
- © 2012 American Heart Association, Inc.
- Tanne D,
- Bates V,
- Verro P,
- Kasner SE,
- Binder JR,
- Patel SC,
- et al
- Grond M,
- Stenzel C,
- Schmulling S,
- Rudolf J,
- Neveling M,
- Lechleuthner A,
- et al
- Saver JL
- Marler JR,
- Jones PW,
- Emr M
- Rose KM,
- Rosamond WR,
- Huston SL,
- Murphy CV,
- Tegeler CH
- Morris DL,
- Rosamond WR,
- Madden K,
- Schultz C,
- Hamilton S
- Lacy CR,
- Suh D-C,
- Bueno M,
- Kostis JB
- Abdullah AR,
- Smith EE,
- Biddinger PD,
- Kalenderian D,
- Schwamm LH
- Katzan IL,
- Graber TM,
- Furlan AJ,
- Sundarajan S,
- Sila CA,
- Houser G,
- et al
- Schwamm LH,
- Fonarow GC,
- Reeves MJ,
- Pan W,
- Frankel MR,
- Smith EE,
- et al
American Hospital Association Guide Book. Chicago, IL: American Hospital Association; 2008.
American Heart Association. Target Stroke. Available at: www.strokeassociation.org/STROKEORG/Professionals/Target-Stroke_UCM_314495_SubHomePage.jsp. Accessed March 16, 2011.
- Reeves MJ,
- Fonarow GC,
- Zhao X,
- Smith EE,
- Schwamm LH
- Reeves M,
- Bhatt A,
- Jajou P,
- Brown M,
- Lisabeth L
- Jones SP,
- Jenkinson AJ,
- Leathley MJ,
- Watkins CL